agentsclimarketplace

Docstrata

Skill linhai0872/docstrata/skill/docstrata

Generate or maintain layered project documentation across knowledge layers (prd, requirements, knowledge, wiki, dev) plus a compact operation to keep long-lived docs bounded. Use when the user asks to document a project, write a PRD / product positioning / roadmap, create a wiki/readme for business people, recover/restore requirements, organize business knowledge materials, write development notes, or compact/shrink bloated docs. Works on any project type (full-stack apps, Dify DSL, CLI/MCP services, skills, or plain document folders).From its SKILL.md

Install
npx -y skills add linhai0872/docstrata --skill docstrata

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 7 stars7 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.8 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

docstrata — 分层知识文档

按知识性质分层,为任意项目生成/维护文档。

wiki / requirements / knowledge / dev 四层对应 CoALA 长期记忆三类(Semantic / Episodic / Procedural)。prd 在这之上,记前瞻产品意图,不属 CoALA 记忆。每层是一个子命令,共享同一套执行循环。compact 是横切收缩操作,把膨胀的层文档收回有界状态。

子命令路由

命令层 / 类型产出何时用
prd产品主张(前瞻 intent)docs/prd.md固化对内的产品定位/价值/功能范围/roadmap
requirementsEpisodicdocs/requirements.md还原/固化需求共识与开发计划
knowledgeSemanticdocs/knowledge/knowledge.md整理业务原始材料为可检索索引
wikiSemanticdocs/wiki.md让所有人(含业务)快速理解整个系统
devProceduraldocs/dev.md记录开发推断与实践结论
index检索入口docs/INDEX.md生成给 coding agent 的文档导航索引
compact横切操作收缩已有层文档层文档随迭代膨胀,手动收回有界状态
all全部五层 + index按依赖顺序一次生成

用户说具体层("生成 wiki"/"还原需求")→ 走对应层。说"全部文档" → all。说"compact/收缩/精简" → compact(可带层名,如 /docstrata compact dev)。不明确则问用户要哪一层。

all 的依赖顺序:prdrequirementsknowledgewikidevindexprd 是最上游意图源头,index 是最下游派生层。跑单层时不强制先跑上游;上游文档已存在则读取引用。

index 详见 references/layer-index.md。纯派生,不走五步循环,但遵循增量更新与 STAMP。

compact 详见 references/compact.md。先压(丢可复原项 + 去重 + 按主题归并)、再拆、归档不删。手动触发,不自动跑。

all 不强制每层都产出。无实质价值的层按 methodology.md 的优雅降级跳过,汇报里说明原因。宁可少而实,勿多而空。

诊断副产物(audit)

任何生成动作完成后,汇总 EXPLORE/MAP 阶段发现的信息质量问题(冲突 / 过期 / 缺口 / 无主孤儿文档),输出信息健康诊断报告。流程副产物,不是单独命令。详见 references/audit-report.md

执行循环(所有层通用)

每层走这五步。方法论细节见 references/methodology.md,执行前务必先读。

  1. EXPLORE — 读现有 docs/ 层文档、README、代码结构、配置。识别项目类型,用对应探索策略。
  2. MAP — 把探索结果映射到本层 completeness contract(见各层 reference),给每个维度打置信度(high/medium/low/missing)。
  3. GRILL — 仅对 low/missing 维度提问。一次一个,附推荐答案。维度全达标则跳过。不重复问已知信息。
  4. GENERATE — 按本层固定骨架生成。已存在则增量更新:diff 现有内容,只重写有实质变化的段落,保留人类手改。
  5. STAMP — 更新 last-verified 日期与变更摘要。

各层契约与骨架(按需加载)

执行某层时,读取对应 reference 获取 completeness contract 和 section 骨架:

执行收缩时:

横切方法论

What ships with it: 11 files

48.4 KB alongside SKILL.md

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.